I'm about to commit several changes to the pmax kernel.
The biggest change is preliminary support for config.new,
(or config as it's now called.)

I believe I've fixed a bug in interrupt initialization on PMAXes
(2100s and 3100s) that would cause these machines to crash if booted
on `busy' Ethernets. Interrupts are now not enabled until
all discovered devices (except SCSI disks and tapes) have been probed
and attached.

I'd like to apologize in advance for any disruption these changes may
cause anyone building NetBSD/pmax kernels I don't have access to
enough machines, or time, to test this as thoroughly as I'd like.
Also, I don't see any way to change the kernel as drastically as this
without causing *some* disruption to those who build kernels.


The changes for new-style config are extensive. Much of the code is
derived from cgd's Alpha port; some is my own invention.  Much of my
code is still rather ugly; but I'm merging them anyway, partly to get
feedback, partly to ``encourage'' testing, and partly because of
other, forthcoming commitments on my time.

Given the benefits that should accrue from using new config, I think
that's the best thing to do.


I've also gone to some effort to ensure that kernels buillt with old
configure will still compile and boot on the machines I have: a
5000/240 and a 5000/200.  Old-config kernels should continue to work
on 3MINS, MAXINES, and PMAXes, but I haven' tested them there.
I am definitely interested in bug reports from these changes.


I'm also in the midst of adding compile-time conditionalization
for 3MIN (5k/1xx) and MAXINE (5k/xx) machines.  These changes are
likely to cause kernel builds to break.  A simple
workaround is to add lines for

	cpu		"DS3100"
	cpu		"DS5000"
	cpu		"DS5000_200"
	cpu		"DS5000_20"
	cpu		"DS5000_100"
	cpu		"DS5000_25"

to your kernel configuration file in sys/arch/pmax/conf, rerun
config.old, and recompile.